New Indexing Method for Content-Based Video Retrieval and Clustering for MPEG Video Database

概要

論文の詳細を見る
Recently, multimedia information, especially video, have vastly increased; as a result, an automatic content-based retrieval became an important research issue in multimedia database. The basic units of a video are a 'frame' and a 'shot'. The results of content-based retrieval is a set of shots that satisfy the user query. In this paper, we introduce an 'average frame' as a logical representative flame of a shot. The average frame is a compact representation of 2 features: color distribution and their locality information of a shot. That would enable users to issue queries such as: select S, from S in shots, where S.match (<locality_info>, <color info>); Here 'shots' is a set of shots in a database, and the predicate 'match' is the operator for retrieval. To retrieve a shot a user has to specify some location, or rather a field of movement which is an area in a frame within which the searched object is moving during the time of a shot, and then specify the amount of color(s) in reference to that specified location.
一般社団法人情報処理学会の論文
1997-09-24